College and Research Libraries Library Growth and Academic Quality THE LATE FREMONT RIDER's famous ob- servations and conclusions on library growth1 were re-examined in a recent article by H. William Axford in this journal.2 Axford sought to find whether or not the growth of university libraries since 1946 was maintaining the pace Rider found typical for the period be- fore 1940, and whether or not Rider's warning that libraries must double in size every fifteen or twenty years in order that their parent institutions might con- tinue to maintain high academic rating is indeed valid. Axford found that uni- versity libraries, in the period 1946-1960, were not growing at a rate which would permit them to double in size every fifteen years, but at only 78 per cent of this rate. He concluded also that "Rider's emphasis on the relationship between the rate of growth of the university li- brary and the over-all quality of the edu- cational program is still essentially cor- rect."3 An examination, however, of the methods whereby Axford's data were chosen, used, and interpreted, may well lead to different conclusions and further questions. It is regrettable that the study of li- brary size and library growth rates is hampered by deficiencies in the quality of available library size and growth sta- tistics. All libraries do not count their holdings in exactly the same way; neither do counting practices in a given library necessarily remain unchanged through- out its history. Progress in intermural 1 Fre mont Rider , "The Gro.wth of American College and University Libraries-and of Wesleya n's" Abo u t B ooks at th e Oli n Library, W esleyan Univer sity, XI (1 940), 1-11. 2 H. William Axford, " Rider Revi s ited, " CRL , XXIII (196 2 ) , 345-47. a Ibid. , p. 347 . MAY 1963 Bv GEORGE PITERNICK Mr. Piternick is Assistant Director of Li- braries at the University of Washington, Se- attle. and intramural standardization has been and is being made; the benefits of each improvement in standardization can be felt, unfortunately, only in statistics gen- erated subsequent to its adoption. Never- theless, library size statistics, if worth gathering . and publishing, are worth using-the degree to which conclusions may be invalidated by flaws in the data must, however, be always considered. American university libraries whose holdings exceeded one million volumes by June 30, 1960, were used by Axford in his study of the first of Rider 's axioms -that college and university libraries, on the average, double in size about every fifteen -:years. Axford's Table 1 illustrates the growth of each of these libraries in the period 1946-1960, ex- . pressing it as percentage increase during the period. Presumably a percentage in- crease of 100 during the period would constitute doubling. Several facts, how- ever, complicate this simple comparison. For one thing, the interval is fourteen years instead of fifteen as Axford as- sumes. Second, the listed library holdings for 1946 are not internally consistent. Some are taken from the July 1947 issue of College and Research Libr~ries~ others from the 1945 edition of the American Library Directory. The figures in the lat- ter publication could apply to the year 1945 at the latest, and frequently refer to a date two or three years earlier. 223 Third, the "average percentage increase" of 78 per cent, as printed, is an un- weighted arithmetic average of individ- ~al percentages and hence of dubious relevance. Last, several errors in listing and computation are detectable. Table I in this paper offers, in revised form, the growth characteristics of the academic libraries Axford selected for study for the period 1946 to 1960. The holdings data for 1946 have been ra- tionalized; they are taken from the July 1947 issue of College and Research Li- braries, supplemented when necessary by data taken from the Princeton Statistics for College and University Libraries for 1945-1946. For each library, the average annual growth rate has been computed as has its corollary value, the doubling time. Doubling time is here used to represent the period of years necessary for a li- brary's holdings to double at the aver- age annual growth rate for the period studied. Table I also shows the growth characteristics of these libraries during the same number of years (fourteen) immediately preceding the dates of Ax- ford's comparison period. For the period 1946 to 1960, the li- braries under study have grown at wide- ly differing rates. Their mean growth rate during this period has been 20.1 years, which, to be sure, indicates a slow- ing down from Rider's average of fifteen years. However, there has been no per- ceptible slowing down in growth rate for these libraries over the previous pe- riod of fourteen years. Their mean dou- bling time between 1932 and 1946 was 20.7 years. Attention is again invited to the wide variation in individual growth rates. In using this ranking as a gauge of over-all academic quality Axford, in his Table 2, did not rank the institutions ac- cording to the Keniston-Berelson rating as he intended, but instead in alphabeti- cal order within two rank clusters used frequently by Berelson; t<?P ten (with the addition of M.I.T. and CalTech) and second ten. Also, Axford's conclu- sion, as mentioned earlier, was evidently arrived at by inspection only. But is Axford's sample typical? His libraries are, after all, the twenty-five largest university libraries, and there- fore not likely typical of all university libraries. Moreover, Axford's sample is not identical with the sample upon which Rider based his conclusions, al- though there are many libraries com- mon to both. A restudy of Rider's sam- ple of twenty libraries (Table 2) shows that their growth rate has definitely de- creased. His first group of ten "represent- ative" large university libraries of re- spectable age he found to have had an average doubling time of sixteen years between 1831 and 1938. Their mean doubling time between 1938 and 1960 was 25.1 years. The second group of ten more recently founded university librar- ies he found to have grown at an aver- age doubling time of 9.5 years (actually 10.9 years) between 1876 and 1938. Their average doubling time was twenty-two years between 1938 and 1960. Again, the individual variation is very large. It appears, then, that any conclusions about the average growth rate of aca- demic libraries must depend largely up- on the libraries chosen for study, and the period during which their growth is studied. Certainly the variation in growth rates is very large. In the twenty- five libraries constituting Axford's sam- ple we find doubling-time values rang- ing all the way from 9.3 years (UCLA, 1932-1946) to 43.7 years (Yale, 1946- 1960). That the growth in size of academic libraries is exponential rather than lin- ear cannot be doubted. Or, put another way, it is clear that every library nor- mally tends to add, during a given year, more volumes than it added during the preceding year and fewer than it will add the next year. It is not clear, how- 224 COLLEGE AND RESEARCH LIBRARIES Library Holdings 19 32 (volumes) Harvard 3,341,700 Yale 2,130,600 Illinois 1,113,733 Columbia 1,358,389 Michigan 839,338 California (Berkeley) 802,817 Cornell 877,393 Chicago 1,012,5·35 Minnesota 682,894 Pennsylvania 773,843 Princeton 678,299 Stanford 567,243 California (Los Angeles) 191,250 Duke 307,601 Northwestern 418,098 Wisconsin 441,500 Ohio 395,725 Texas 436,224 Indiana 247,320 Johns Hopkins 423,501 New York . Washington (Seattle) 316,136 Brown 430,683 Iowa 352,192 Missouri 292,268 Mean 767,970 TABLE 1 RECENT GROWTH OF UNIVERSITY LIBRARIES (after Axford, 1962) Average Annual Doubling Holdings Growth Time 1946 1932-46 1932-46 (volumes) (per cent) (years) 4,804,968 2.6 27.0 3,5 39,596 3.7 19.1 2,003,622 4.3 16.5 1,778,058 1.9 36.8 1,267,518 3.0 23.4 1,378,602 3.9 18.1 1,206,19·5 2.3 30.5 1,584,264 3.2 22.0 1,422,529 5.4 13.2 1,033,794 2.1 33.3 1,058,920 3.2 22.0 952,051 3.8 18.6 543,281 7.7 9.3 740,493 6.5 11.0 788,832 4.6 15.4 600,000 2.2 31.9 709,875 4.3 16.5 801,637 4.4 16.1 617,947 6.8 10.5 737,760 4.0 17.7 715,157 594,320 4.6 15.4 665 ,041 3.2 22.0 665,930 4.7 15.1 525,557 4.3 16.5 1,229,438 3.4 20.7 Average Annual Doubling Holdings Growth Time 1960 1946-60 1946 -60 (volumes) (per cent) (years) 6,697,111 2.4 29.2 4,394,988 1.6 43.7 3,28t3,158 3.6 19.6 2,875,761 3.5 20.1 2,818,341 5.9 12.1 2,503,060 4.4 16.1 2,116,230 4.1 17.3 2,094,824 2.0 35.0 1,968,101 2.3 30.5 1,665,114 3.5 20.1 1,626,537 3.1 22.7 1,592,287 3.7 19.1 1,464,308 7.3 9.8 1,435,164 4.8 14.8 1,429,431 4.3 16.5 1,384,222 6.2 11.5 1,369,348 4.8 14.8 1,350,671 3.8 18.6 1,317,269 5.6 12.7 1,159,747 3.3 21.3 1,067,946 2.9 24.2 1,060,086 4.2 16.8 1,025,479 3.1 22.7 1,021,441 3.1 22.7 1,002,263 4.7 15.1 1,989,115 3.5 20.1 Library Holdings 1831 (volumes) Harvard 39,605 Yale 25,500 Columbia 4,580 Princeton 8,000 Penns ylvania 2,000 Brown 11 ,662 North Carolina 4,800 Virginia 8,000 Rutgers 6,500 Georgetown 7,000 Mean I 11 ,764 Library Holdings 1876 (volumes) Chicago 18,000 California (Berkeley) 13,600 Illinois 11 ,000 Cornell 39,000 Minnesota . . 10,000 Western Reserve 10,000 Iowa 8,823 Oberlin 14,000 Rochester 12,000 Syracuse 10,000 Mean 14,642 TABLE 2 GROWTH OF COLLEGE AND UNIVERSITY LIBRARIES (after Rider, 1940) a) Represe ntative Large University Libraries Average Annual D oubling Holdings Growth Time 1938 1831- 1938 1831-1938 (volumes) (per cent) (years) 3,941,359 4.4 16.1 2,748,000 4.5 ° 15 .7 1,615 ,051 5.6 12.7 919,555 4.5 15.7 881,781 5.9 12.1 530,290 3.6 19.6 357,629 4.1 17 .3 303,502 3.5 20.1 273,873 3.6 19.6 258 ,700 3.4 20 .7 1,182,974 4.4 16.1 b) Recently Founded University Libraries Average Annual Doubling Holdings Growth Time 1938 1876-1938 1876- 1938 (volumes) (per cent) (years) 1,232,745 7.1 10.1 1,141 ,612 7.2 10.0 1,130,000 7.8 9.2 1,035,170 5.4 13.2 I ,017,690 7.7 9.3 508,000 6.5 11.0 441,396 6.5 11.0 386,664 5.5 12.9 329,700 5.5 12.9 313,454 5.7 12.5 753,643 6.6 10.9 L---------------------------------------------------~----------------------- Average Annual Doubling Holdings Growth Time 1960 1938 -60 1938-60 (volumes) (per cent) (years) 6,697,111 2.4 29.2 4,394,988 2.2 31.9 2,875,761 2.7 26.0 1,626,5 37 2.6 27.0 1,665 ,114 2.9 24.2 1,025,479 3.0 23 .4 1,025,944 4.9 14 .5 966,390 5.4 13 .2 907,452 5.6 12 .7 399,567 2.0 35.0 2,158,434 2.8 25.1 Average Annual Doubling Holdings Growth Time 1960 1938-60 1938-60 (volumes) (per cent) ( years) 2,094,824 2.4 29.2 2,503,060 3.6 19.6 3,288,158 5.0 14:2 2,116,230 3.3 21.3 1,968,101 3.0 23.4 755,000 1.8 38.8 1,021,441 3.9 18.1 544,494 1.6 43.7 696,630 3.5 20.1 522,549 2.3 30.5 1,510,487 3.2 22.0 ever, that the rate of growth for .a given library does or should remain invariable throughout the library's existence, or that the growth rates of all comparable libraries should be the same or closely similar, considering the great number of internal and external educational and economic factors affecting library acqui- sition. The commonly quoted values of fifteen , sixteen, or twenty years for "nor- mal" or "average" or "typical" doubling times need not be accepted as such. In- deed, there appears to be no obvious reason why Rider considered fifteen or sixteen years typical. On the basis of his own computations of value thirteen years for doubling would be more typical for the twenty "university" libraries he stud- ied. The significance of the growth rate as a meaningful datum remains to be estab- lished. Rider's state.ment that " ... we may assert this as almost axiomatic: un- less a college or university is willing ... not to maintain its place in the steady flow of cultural development, it seems to be inevitable that it must double its li- brary in size every fifteen or twenty years" 4 seems to establish rate of growth as a datum independent of the basis up- on which this growth occurs, i.e.) the ab- solute size of the holdings. That is, it doesn't matter how large a library is, as long as it grows rapidly. As the size of the library increases, the annual increments necessary to support <;t rate of growth which will result in doubling every fif- teen or twenty years become very large, and, if projected very 'far, become astro- nomical in size. This fact very likely caused Metcalf, in 1954, to say, "When a library has reached maturity ... if its book collections increase very much more than 21;2 per cent a year, the library is growing more rapidly than it should."5 4 Rider, op. cit., p. 11. 5 K. D. Metcalf, "Spatial Problems in University Libraries" Library Trends, Il ' (1954), 559. MAY 1963 That is, if it is large enough, it need not and should not grow rapidly. But these contradictory quotations are, after all, subjective evaluations, al- though made, to be sure, by eminent and respected authorities. Axford has used a more objective measure of the rele- vance of a library's growth rate to . the academic quality of its parent institu- tion. Using an available academic rank- ing of institutions offering graduate study in many fields Axford concluded that "Rider's emphasis on the positive relationship between the rate of growth of the university library and the over- all quality of the educational program is still essentially correct."6 Again, ques- tionable selection, listing, and analysis of the data appear to have lead to de- batable conclusions. Dean Haywood Keniston, in the course of an educational survey of the Univer- sity of Pennsylvania7 constructed a rank order of universities based upon the ex- cellence of their graduate teaching and research programs in the arts ·and sci- ences, as rated by a large number of graduate deans and departmental chair- men in their specialities throughout the nation's leading universities. This rat- ing was accepted by Bernard Berelsons as being essentially sound, and was used by him in many of the comparisons in his work, although he added the Cali- fornia Institute of Technology and the Massachusetts Institute of Technology to the top twenty schools, for purposes ~ot. relevant here. Keniston's study was hmited to those institutions offering graduate study in a wide variety of fields. If, however, certain rank orders of library data for the twenty institutions (omitting M.I.T. and CalTech because they have academic offerings in only a 6 Axford, op . cit., p. 347. . 7 Haywood Kenis~on, Graduate . Study and Research m t~e Arts f!nd Sclf!nces at _the University of Pennsyl- van~a. (Philadelphia: Umve rsit y of Pennsylvania Press, 1959). 8 _Be rn ard Berelson, Gradt{ate Education in the Umted States. (New York: McGraw-Hill, 1960). 227 TABLE 3 CoRRELATIONS AMONG UNIVERSITY AcADEMic RANKING, LIBRARY HOLDINGS, AND GROWTH RATE LIBRARY LIBRARY AVERAGE ANNUAL HOLDINGS HOLDINGS GROWTH RATE Keniston- 1946 INSTITUTION Berelson Rank Million Order Vo.lumes Top ten: 4.80 Harvard I California (Berkeley) 2 1.38 Columbia 3 1.78 Yale 4 3.54 Michigan 5 1.27 Chicago 6 1.58 Princeton . 7 1.06 Wisconsin 8 0.60 Cornell 9 1.21 Illinois 10 2.00 Second ten: Pennsylvania 11 1.03 Minnesota 12 1.42 Stanford 13 0.95 California (Los Angeles) 14 0.54 Indiana 15 0.62 Johns Hopkins 16 0.74 Northwestern 17 0.79 Ohio I8 0.71 New York 19 0.72 Washington (Seattle) 20 0.59 Correlation Coefficient with Keniston-Berelson Rank Order limited number of fields and relatively small libraries) are compared with the actual Keniston-Berelson rank order (Table 3) a different conclusion must be reached. The correlation between Keniston-Berelson rank order and li- brary growth rate rank order during the last fourteen years (r = -.24) is not sig- nificant. On the other hand, the correla- tion between the Keniston-Berelson rank order and the rank order of absolute size of library holdings is highly signifi- cant for 1946 hpldings (r = .76), and even more spectacula~ly significant for 1960 holdings (r = .87). Berelson reported that he had en- countered, in interview, some dispute as to the validity of the Keniston aca- demic ranking in some specific areas. 1960 1946-60 Rank Million Rank Rank Order Vo.lumes Order Per Cent Order I 6.70 I 2.4 I7 7 2.50 6 4.4 6 4 2.88 4 3.5 I2.5 2 4.39 2 1.6 20 8 2.82 5 5.9 3 5 2.09 8 2.0 I9 10 1.63 11 3.1 15 18 1.38 15 6.2 2 9 2.12 7 4.1 9 3 3.29 3 3.6 11 11 1.67 10 3.5 12.5 6 1.97 9 2.3 18 12 1.59 12 3.7, 10 20 1.46 13 7.3 1 17 1.32 17 5.6 4 14 1.16 18 3.3 14 13 1.43 14 4.3 7 16 1.37 16 4.8 5 15 1.07 19 2.9 16 19 1.06 20 4.2 8 .76 .87 -.24 Answering these en ticisms, he reports, would result in moving a given institu- tion one or at most two places up or down on the list, and this would make very little difference in the values of the coefficients of correlation. Whatever crit- icisms of the detailed ranking he en- countered were obviated by using the ranking in the two clusters of top ten and second ten, and it is in this form that Berelson used it in most of his analyses. He reports almost complete agreement with the ranking as used in this way. Using such clusters in comparing li- brary size and growth data with aca- demic ranking (Table 4) leads to the same conclusions as does the calculation of coefficients of correlation. The mean 228 COLLEGE AND .- RESEARCH LIBRARIES TABLE 4 LIBRARY HOLDINGS AND GROWTH OF GRADUATE STUDY INSTITUTIONS Mean Keniston- Holdings Berelson 1946 Ranking (volumes) Top Ten . 1,922,174 Second Ten 811,555 Other AGS Members 523,640 of the top ten institutions had over twice as many volumes in 1946 and 1960 as did the mean of second ten institutions and over three times as many volumes as the mean of unranked other institutions belonging to the Association of Graduate Schools. 9 During the period of study the mean of top ten institutions added al- most twice as many volumes as the mean of second ten institutions and almost three times as many as the mean of other unranked AGS members. At the same time, the mean growth rate of the lat- ter two clusters was significantly higher than the growth rate of the mean ' top ten institutions. This comparison is probably even more meaningful than is the comparison of rank order correlations. Not only does it avoid the minute determination and comparison of library holdings, data which are somewhat unreliable, but of even greater importance, it shows that the magnitude of yearly gross holdings additions, like the holdings themselves, are of more significance than growth rate in determining library quality. That the absolute size of a university library's holdings and the absolute size of its yearly gross increments, and not its current growth rate, are the best measures of its quality has other con- firmation, as Fussier has pointed out.10 Those large university libraries, notably 9 Relevant size figu.res were readily av'ailable for only fourteen of these unranked institutio.ns. 10 H. H. Fussier, personal communication, October 5, 1962. MAY 1963 Mean Mean Mean Annual Mean Holdings Additions Growth Rate Doubling Time 1960 1946-60 1946-6 0 1946-60 (volumes) (volumes) (per cent) (years) 2,979,923 1,057,749 3.2 22.0 1,409,364 597,809 4.0 17.7 905 ,607 381,967 4.0 17.7 Chicago and Yale, which are carrying out weeding programs, adversely affect their growth rates thereby. At the same time, they are undoubtedly improving their libraries and they remain in the upper ranks of research institutions in spite of having very low growth rates, because they remain large libraries add- ing large yearly increments to their hold- ings. On the basis of growth rates alone they would be considered to be in sad decline. Strong positive correlation between two sets of data does not in itself estab- lish causal connection between them. The existence of high correlation be- tween absolute size of library holdings and the academic quality of an institu- tion's graduate program does not estab- lish that the former causes the latter any more than that the latter causes the for- mer. It does suggest strongly, however, that the two are not independent-very likely cause and effect are intertwined, as Clapp has concluded. "And just as good graduate students are attracted to good teachers, so the good scholars are attracted to institutions having good li- braries. And-and this is the whole point-libraries which are good for the diversity of interests which are repre- sented in a university faculty necessarily are or soon become large libraries.11 •• 11 Verner Clapp, "Graduate Education and Library Resources," Journal of the Graduate Research Center (of Southern Methodist University), XXX (1962), 51. 229